Fuzzy ARTVar : an improved fuzzy ARTMAP algorithm

Abstract

We introduce a variation of the performance phase of fuzzy ARTMAP which is called Fuzzy ARTVar. Experimental results have shown that Fuzzy ARTVar exhibits superior generalization performance, compared to fuzzy ARTMAP, for a variety of machine learning databases. Furthermore, experimental results have also demonstrated that Fuzzy ARTVar compares favourably with other existing variations of fuzzy ARTMAP, such as ARTEMAP (power rule), ARTEMAPQ (Q-max rule), and Gaussian ARTMAP. The performance of Fuzzy ARTVar is independent of the tuning of network parameters, which is in contrast with the ARTEMAP, ARTEMAPQ, and Gaussian ARTMAP algorithms, whose performance depends on the choice of certain network parameters.